I have a test case quicksave.sfs and a few screenshots showing the ascent settings and throttle sputtering here: https://dl.dropboxusercontent.com/u/8244638/quicksave.zip

In this example case there's plenty of thrust to still get to orbit despite the problem. But in a lot of my other craft with worse TWR, this has been fatally killing my takeoffs from Mun, Minmus, etc. I prefer very low takeoff paths for efficiency when there's no atmosphere, this may be bringing out the problem more easily?

In 2.0.6, it works fine. 2.0.7 did something, my guess is one of 14e89d1 or 4875049 or 15ebe70

It happens whether or not corrective steering is enabled. The steering error jumps around a lot when the throttle is sputtering. Note that this happens before apoapsis gets particularly close to the target orbit altitude, you can ask for higher orbits and still see it.

Yes, the problem is definitely with the code from 15ebe70. It looks like maybe when the orbit becomes nearly circular, the calculation runs into numerical precision problems. I'll figure something out. Thanks @tavert for the excellent test case.

I just tried this with corrective steering on and off. With corrective steering off, the sputtering was greatly reduced, and I ended up with a 10 km apoapsis and a total of 615.4 m/s expended (counting the circularization burn, which I didn't execute). With CS on, the sputtering was very noticeable. My resulting apoapsis was 12.2 km, and the total delta-V would have been 629.4 m/s after circularization.

I'm suspicious of 15ebe70 (and not just because it's the only one of those three commits that I didn't write =)). @Anatid, any thoughts?